Title of Paper: Scalability Study of the Ad hoc On-Demand Distance Vector Routing Protocol

Authors: Sung-Ju Lee, Elizabeth M. Belding-Royer, and Charles E. Perkins

Your Name: Khashayar Dehdashtinejad

What are the most important points made in the paper?

1-Clustering routing protocols group nodes into clusters based on their proximity to each other. Each cluster generally has a cluster leader, which is the representative of the nodes in its cluster. The cluster leader typically participates in the network routing protocol, freeing the other network nodes of this burden. Routes in clustered networks may often be recorded hierarchically between clusters. Clusters require periodic messaging from each network node in order to maintain the clusters. This periodic messaging results in higher processing and control packet overhead, as well as increased bandwidth utilization and longer delays. Moreover, if the protocol constrains routes to traverse cluster leaders, longer path lengths will be required. This is another contributor to increased bandwidth utilization. Finally, there may be complications when the cluster leaders fail or give up their cluster leader status.

2- Core Extraction Distributed Ad Hoc Routing (CEDAR) algorithm. CEDAR is an algorithm that builds a set of nodes to perform route computation. Using the local state information, a minimum dominating set of the network is approximated to form the core. CEDAR establishes QoS routes that satisfy bandwidth requirements using the directionality of the core path. Link state and bandwidth availability is exchanged to maintain important information for computing QoS routes. Although CEDAR builds a core infrastructure that yields low overhead, the protocol is fairly complex and difficult to implement. The problem of calculating the minimum dominating set and the core is known to be NP-hard.

3- On demand routing protocols like RDMAR and ZRP suffer from the same disadvantage as the class of on-demand protocols whereby efficiency drops as the number of source-destination pairs increases, due to the likely requirement of a route discovery flood.

4- Ad hoc On-Demand Distance Vector (AODV) is an on-demand protocol which is capable of providing unicast, multicast, and broadcast communication.  AODV utilizes per node sequence numbers. A node increments the value of its sequence number whenever there is a change in its local connectivity information.

5- Route discovery begins when a source node needs a route to some destination. It places the destination IP address and last known sequence number for that destination, as well as its own IP address and current sequence number, into a Route Request (RREQ). It then broadcasts the RREQ and sets a timer to wait for a reply. When a node receives the RREQ, it first creates a reverse route entry for the source node in its route table. It then checks whether it has an unexpired route to the destination node. In order to respond to the RREQ, the node  must either be the destination itself, or it must have an unexpired route to the destination whose corresponding sequence number is at least as great as that contained in the RREQ. If neither of these conditions are met, the node rebroadcasts the RREQ. Route Reply (RREP) places the current sequence number of the destination, as well as its distance in hops to the destination, into the RREP, and then unicasts this message back to the source. The node from which it received the RREQ is used as the next hop. When an intermediate node receives the RREP, it creates a forward route entry for the destination node in its route table, and then forwards the RREP to the source node.

6- The RERR contains the IP address of each destination which is now unreachable, due to the link break. The RERR also contains the sequence number of each such destination, incremented by one. The node then broadcasts the packet and invalidates those routes in its route table.

7- The expanding ring search and query localization can be used to reduce the area searched during a route discovery, and hence prevent flooding of the network. The current Internet draft specification of AODV recommends such an expanding ring search be used for route discoveries. Local repair can also be used to provide immediate patching of breaks in active routes.  An expanding ring search works by searching successively larger areas, centered around the source node, until a node with a route to the destination is located. The basic premise behind the expanding ring search is to find some local node with a route to the destination, and thereby avoid flooding the entire network in search of such a route.

8- Query localization is a method by which the flooding of the RREQ is restricted to some area that is based on the previously known route to the destination node. Hence the RREQ is not actually flooded at all, but instead is limited to a specific region of the network. As with the expanding ring search, the drawback of the query localization protocol occurs when a route to the destination is not located on the first attempt. This results in certain nodes being repeatedly queried, as well as an increased route acquisition latency.

9- It is interesting to note that AODV-QL has the poorest throughput. The main purpose of query localization is to exploit node locality and reduce the number of routing message transmissions. Localizing the query, however, has the risk of not being able to establish the route.

10- It is observed that schemes that utilize the local repair technique yield longer paths. For protocols that do not use local repair, only the source node can reconstruct routes. When a source rediscovers routes with a request/reply cycle, a new route is obtained based on current network information such as hop count, route freshness, node location, network topology, etc. On the other hand, in local repair schemes, the node immediately upstream of the disconnected link initiates a route reconstruction. Because of the possibility that the destination has actually moved closer to the source node, but the distance between the node reconstructing the route and destination has increased, path lengths tend to grow as intermediate nodes repair routes.

11- Schemes that utilize the local repair technique have shorter delays. Protocols in which sources initiate route repair have longer end-to-end delays because of longer route re-establishment latency. To recover a broken route, a RERR packet must first be delivered from the node upstream of the broken link to the source of the route. The RREQ must then be broadcast from the source to the destination, and a RREP consequently has to be transmitted back to the source.

What are the main weaknesses of the paper? Do you believe the assumptions and results? If not, why not?

Some of assumptions in simulation is not clear enough. For example, why in trafic generator the size of data payload choose to be 512 bytes?

If possible, compare this work with other work that you have heard about. If you chased down additional information on this topic, what was it?


List three questions or points that you plan to bring up in the discussion. 
1. More information about routing on demand, distance vector and link state.
2. Query localization is still not clear to me
3. Why ad hoc network performance is optimal when the average number of neighbors is between six and eight?

Place a number for each item below corresponding to your evaluation.

Scale: 1 poor, 2 fair, 3 medium, 4 high, 5 outstanding.

Quality of presentation (writing, exposition, etc.): 4

Technical contribution: 4

Novelty of ideas: 3

Overall: 4

http://www.hpl.hp.com/personal/Sung-Ju_Lee/abstracts/papers/ijnm2002.pdf

Title of Paper:  Beyond Interoperability – Pushing the Performance of Sensor

Network IP Stacks

Authors: JeongGil Ko, Joakim Eriksson, Nicolas Tsiftes, Stephen Dawson-Haggerty, Jean-Philippe Vasseur, Mathilde Durvy, Andreas Terzis, Adam Dunkels and David Culler

Your Name: Khashayar Dehdashtinejad

What are the most important points made in the paper?
1- RPL supports multi-topology routing: a set of virtual routing topologies that can be built that are optimized for different metrics and sets of constraints.

2- RPL attempts to support different application requirements through multiple objective functions (OFs), used for parent selection. These OFs are carried in DIOs and describe how a node should compute its DODAG rank value given a set of metrics. An objective function can be tailored so that RPL computes routing paths that achieve application specific goals.

3- 6LoWPAN provides a link-layer fragmentation and reassembly mechanism, also provides Header compression which is a well-known technique to reduce the header size of IP packets.

4- Different MAClayer retransmission intervals and message queue sizes between the two implementations cause the performance degradation in mixed networks.

5- The end-to end number of transmission attempts of Contiki Collect is the most efficient. Because Contiki Collect incorporating an active probing scheme for precise link quality estimations. While the four-bit link estimator that TinyOS CTP uses relies only on beacon sequence numbers to estimate ETX between nodes that do not exchange data packets, Contiki Collect sends explicit unicast probe messages to the neighbors that lack link quality samples when computing the ETX. This allows Contiki Collect to make more frequent and accurate link quality measurements thereby promptly discovering paths that can deliver packets with lower costs.

What are the main weaknesses of the paper? Do you believe the assumptions and results? If not, why not?

- Lots of references to standard groups and acronym names without enough explanation makes this paper hard to read.

If possible, compare this work with other work that you have heard about. If you chased down additional information on this topic, what was it?


List three questions or points that you plan to bring up in the discussion. 
1.  More information about Destination-Oriented Directed Acyclic Graph (DODAG)

2.  What id RPL routing protocol?

3.  What is power-line communication (PLC)?

4- What is Point-to-Point Protocol (PPP) and Serial Line IP (SLIP)?

5- More information about ETX (to estimate link quality) and EWMA filter

6- What is a unit disk graph model with Bernoulli losses?

Place a number for each item below corresponding to your evaluation.

Scale: 1 poor, 2 fair, 3 medium, 4 high, 5 outstanding.

Quality of presentation (writing, exposition, etc.): 2

Technical contribution: 1

Novelty of ideas: 2

Overall: 1

https://www.sics.se/~adam/ko11beyond.pdf

Title of Paper: Lossy Links, Low Power, High Throughput

Authors: Simon Duquennoy, Fredrik Osterlind, Adam Dunkels

Your Name: Khashayar Dehdashtinejad

What are the most important points made in the paper?

1- burst forwarding, a generic packet forwarding technique that combines low power consumption with high throughput for multi-purpose wireless networks. Burst forwarding uses radio duty cycling to maintain a low power consumption, recovers efficiently from interference, and inherently supports both single streams and cross-traffic.

2- Burst forwarding groups multiple packets into bursts that are transmitted over each hop with a stubborn link layer performing retransmissions on millisecond time scales. Bursts are coordinated across multiple hops using a two-level retransmission scheme and multi-channel operation. The nodes use radio duty cycling to keep the radio off more than 99% of the time. They wake up periodically to receive long bursts of data, store them in flash memory, and forward them to the next hop.

3- Two-level retransmission, rapid link-layer retransmissions—in combination with less rapid MAC-layer retransmissions.  The aim of the link-layer retransmissions is to recover from isolated losses and high-frequency radio interference. When no acknowledgment is received within a given fixed period, the current burst is stopped and the second layer of retransmissions is involved. The second level of retransmissions operates at the MAC layer, at the granularity of bursts rather than individual packets. It consists of a classical CSMA with increasing backoff, which is known to provide a good trade-off between energy efficiency and latency.

4- burst forwarding uses multiple channels to avoid both intra-path interference between nodes involved in the same transfer and inter-path interference between nodes or networks having independent activities. This mechanism allows burst transmissions to be pipelined over multiple hops. When the density of the network allows multi-channel to avoid all intra-path interference, a full pipeline takes place involving all nodes in the forwarding path: node n receives a burst from n1, then it forwards the enqueued data to n+1 while n-1 acquires new data.

What are the main weaknesses of the paper? Do you believe the assumptions and results? If not, why not?

1-The fact that they save bursts in flash and they need to fragment and reassemble packets in source and destination only can not be useful if every node need to react to some event (1 to many relationship)

2- The first transmission is very costly because of multiple channel and wakeup ack design.

3- Burst forwarding by itself is not 100% reliable.

4- I’m not sure what happens if write or read to flash memory fails.

If possible, compare this work with other work that you have heard about. If you chased down additional information on this topic, what was it?

I think RCRT is better protocol than burst forwarding since it has rate control mechanism, and congestion control.

List three questions or points that you plan to bring up in the discussion. 
1. More information about Flush and PIP protocols.
2. Why link layer acknowledgment needed for wakeup transmission period?
3.How they guarantee atomic retransmission?

4- What is CSMA with increasing backoff?

Place a number for each item below corresponding to your evaluation.

Scale: 1 poor, 2 fair, 3 medium, 4 high, 5 outstanding.

Quality of presentation (writing, exposition, etc.): 4

Technical contribution:  4

Novelty of ideas: 4

Overall: 4

http://www.sics.se/~adam/duquennoy11lossy.pdf

Good notes: TCP was designed under the loss assumptions in the traditional wired Internet and is known to be ill-suited for use over lossy links [2] and wireless networks. There are at least four reasons for this: TCP treats packet loss as congestion, which leads to suboptimal throughput over lossy links; TCP headers are large, which is a problem over low-bandwidth links; TCP uses positive acknowledgments, which may overwhelm the wireless medium; and TCP timing may interfere with link-layer recovery mechanisms.

Title of Paper: RCRT: Rate-Controlled Reliable Transport

for Wireless Sensor Networks

Authors: Jeongyeup Paek, Ramesh Govindan

Your Name: Khashayar Dehdashtinejad

What are the most important points made in the paper?
Come up with at lease three. 

1- using centralized end-to-end delivery protocol that do the congestion control in the sink and can be adopt by existing system with not changing the node’s code.

2- nodes can have multiple applications running at same time and send those to sink concurrently.

3- RCRT provides 1- reliable end-to-end transmission 2-network efficient with avoiding congestion collapse 3- support concurrent applications 4- it is flexible with capacity allocation policy 5- minimal sensor functionality 6- robustness to routing dynamic

4- RCRT hast three main components. 1-Congestion detection 2- rate adoption (total sustainable trafic ?? in network ) 3- rate allocation  

5- using time to recover loss as a congestion indicator

What are the main weaknesses of the paper? Do you believe the assumptions and results? If not, why not?

1- the assumption that there is only a sink in the system may not be valid in some applications.

2- the fact that each node should store a copy of every data packet may not be feasible.

3- the fact that it is centralized system can have a potential single point of failure.

If possible, compare this work with other work that you have heard about. If you chased down additional information on this topic, what was it?


List three questions or points that you plan to bring up in the discussion. 
1. what is exactly IFRC?
2.  What is RBC?
3.  What is NACK?

4- what will happen if we have multiple sinks?

5- how does A and M(t) are calculated in rate adaption formulas?
Place a number for each item below corresponding to your evaluation.

Scale: 1 poor, 2 fair, 3 medium, 4 high, 5 outstanding.

Quality of presentation (writing, exposition, etc.):4

Technical contribution: 4

Novelty of ideas: 4

Overall:4

Title of Paper: Anatomizing Application Performance Differences on Smartphones

Authors: Junxian Huang, Qiang Xu, Birjodh Tiwana, Z. Morley Mao, Ming Zhang, Paramvir Bahl

Your Name: Khashayar Dehdashtinejad

What are the most important points made in the paper?

1- carefully measure network performance such as TCP throughput, downlink RTT, retransmission rate, Local DNS lookup time, TCP handshake time, Ping latency, controlled measuring of impact of network, impact of concurrency, impact of compression, impact of javascript execution speed, impact of data URL schema

2- more concurrent TCP connections can potentially increase web browsing performance

3- Content compression can improve web performance irrespective of RTT

4- it helps to decrease TCP idle time by having mobile version of website with less Javascript objects.

What are the main weaknesses of the paper? Do you believe the assumptions and results? If not, why not?

Some of experiments result such as Time of day correlation may not be accurate enough because they run the experiment in only their place and not several places.

If possible, compare this work with other work that you have heard about. If you chased down additional information on this topic, what was it?


List three questions or points that you plan to bring up in the discussion. 
1. What is chatty protocol?
2. What is Data-URL schema?
3. How network carriers can customize videos base on network condition?

Place a number for each item below corresponding to your evaluation.

Scale: 1 poor, 2 fair, 3 medium, 4 high, 5 outstanding.

Quality of presentation (writing, exposition, etc.):

Technical contribution: 3

Novelty of ideas:  2

Overall: 3

http://sing.stanford.edu/cs303-sp11/papers/r16-smartphones.pdf

Title of Paper:  The Hitchhiker’s Guide to Successful Wireless Sensor Network Deployments

Authors: Guillermo Barrenetxea, François Ingelrest, Gunnar Schaefer, and Martin Vetterli

Your Name:  Khashayar Dehdashtinejad

What are the most important points made in the paper?
1- synchronization which syncs the clock of all motes with server to provide correct absolute time. 

2- Remote control and Don’t be black box. When I think about outdoor deployment noting is more important of remote control. 

3- Data you can Trust. I think is really important to analyse the data you gathered to verify it’s validity.

4- layers of their network. 1-Application 2- Transport 3- Network 4- MAC

What are the main weaknesses of the paper? Do you believe the assumptions and results? If not, why not?

This paper explains the experiences of several WSN deployment. It has valuable insight over real world deployment but it lacks more explanation about the technologies they use. Such as Tiny OS, SensorScope, TinyNode, WSim, GNU Radio, SHT75 and a lot more can be talked more in depth.

If possible, compare this work with other work that you have heard about. If you chased down additional information on this topic, what was it?


List three questions or points that you plan to bring up in the discussion. 
1. What is difference between ulti-hop and single-hop?
2.  It isn’t clear that what author meant by publish early?
3.  More clarification about tagging data with version data.

Place a number for each item below corresponding to your evaluation.

Scale: 1 poor, 2 fair, 3 medium, 4 high, 5 outstanding.

Quality of presentation (writing, exposition, etc.): 4

Technical contribution: 3

Novelty of ideas: 4

Overall: 4

http://infoscience.epfl.ch/record/125712/files/2008-sensys.pdf

Title of Paper:  THALES RESEARCH AND TECHNOLOGY - Wireless Sensor Networks

Authors:  Tim Baugé

Your Name:  Khashayar Dehdashtinejad

What are the most important points made in the paper?

1. Self-organisation algorithms and protocols
2. Low power WPAN wich requires a node to compress/decompress and fragment packets
3. WirelessHART integrating HART protocol with wireless bearers.  

What are the main weaknesses of the paper? Do you believe the assumptions and results? If not, why not? 

I think it is brief overview of new technologies in the area of wireless sensor network but it doesn’t give a reader enough knowledge to understand those technologies.

If possible, compare this work with other work that you have heard about. If you chased down additional information on this topic, what was it?


List three questions or points that you plan to bring up in the discussion. 
1.  What is difference between Zigbee and WirelssHart?
2.  What is self-organizing algorithm and protocol?
3.  What we can learn from RFID technology to adopt in wireless network sensor?

Place a number for each item below corresponding to your evaluation.

Scale: 1 poor, 2 fair, 3 medium, 4 high, 5 outstanding.

Quality of presentation (writing, exposition, etc.):

Technical contribution: 2

Novelty of ideas: 2

Overall: 2

Link to paper: http://www.thalesresearch.com/Publications/WhitePapers/Documents/NET090901.pdf

—————————

Title of Paper:  Programming Wireless Sensor Networks: Fundamental Concepts and State of the Art

Authors:  Luca Mottola and Gian Pietro Picc

Your Name:  Khashayar Dehdashtinejad

What are the most important points made in the paper? 

1. A taxonomy of WSN applications (sense only, sense and react, interaction pattern, space and time)
2. MAC (Medium Access Control) protocol for efficient communications. Contention based protocols and time-slotted protocols. 
3. OS and system services for WSN

What are the main weaknesses of the paper? Do you believe the assumptions and results? If not, why not?
More information about power source and their rules on WSN would be great i.e power loss, different types of sources. Also I was looking forward to get more information on routing, socket programing and sensor programing.

If possible, compare this work with other work that you have heard about. If you chased down additional information on this topic, what was it?


List three questions or points that you plan to bring up in the discussion. 
1.  Why dynamic linking is important for reprograming?
2.  How Contention based protocols actually implemented?
3.  What is ZerbraNet project?

Place a number for each item below corresponding to your evaluation.

Scale: 1 poor, 2 fair, 3 medium, 4 high, 5 outstanding.

Quality of presentation (writing, exposition, etc.):

Technical contribution: 4

Novelty of ideas: 3

Overall:4